From mboxrd@z Thu Jan 1 00:00:00 1970 From: mst@redhat.com (Michael S. Tsirkin) Date: Mon, 14 Oct 2013 13:46:04 +0300 Subject: [PATCH 0/3] virtio-mmio: handle BE guests on LE hosts In-Reply-To: <1381741495.3247.18.camel@hornet> References: <1381502171-8187-1-git-send-email-marc.zyngier@arm.com> <20131012182837.GC2089@redhat.com> <525BAA40.8070103@arm.com> <20131014085931.GB30954@redhat.com> <1381741495.3247.18.camel@hornet> Message-ID: <20131014104604.GC30954@redhat.com> To: linux-arm-kernel@lists.infradead.org List-Id: linux-arm-kernel.lists.infradead.org On Mon, Oct 14, 2013 at 10:04:55AM +0100, Pawel Moll wrote: > On Mon, 2013-10-14 at 09:59 +0100, Michael S. Tsirkin wrote: > > On Mon, Oct 14, 2013 at 09:24:32AM +0100, Marc Zyngier wrote: > > > Hi Michael, > > > > > > On 12/10/13 19:28, Michael S. Tsirkin wrote: > > > > On Fri, Oct 11, 2013 at 03:36:08PM +0100, Marc Zyngier wrote: > > > >> This small patch series adds just enough kernel infrastructure and > > > >> fixes to allow a BE guest to use virtio-mmio on a LE host, provided > > > >> that the host actually supports such madness. > > > >> > > > >> This has been tested on arm64, with some fixes to KVM and a set of > > > >> changes to kvmtool, both which I am posting separately. > > > >> > > > >> A branch containing all the relevant changes is at: > > > >> git://git.kernel.org/pub/scm/linux/kernel/git/maz/arm-platforms.git kvm-arm64/be-on-le-3.12-rc4 > > > >> > > > >> Cc: Rusty Russell > > > >> Cc: Michael S. Tsirkin > > > >> Cc: Pawel Moll > > > > > > > > We are changing the spec to make everything LE instead of > > > > the native endian. > > > > > > > > I think that'll fix the issue in a cleaner way. > > > > > > While I agree that it would solve the issue completely, it would also > > > break all BE users. Is that really an option? > > > > I proposed several ways to create "transitional devices" > > which can detect and switch to old interface at run-time. > > > > Pawel thinks that's not necessary so ... > > Don't wipe yourself with my name, please. > > You forgot to mention that the devices are versioned and the behavior of > the legacy devices remains unchanged. No existing implementation will be > broken. > > Pawe? > I'm sorry if what I wrote was misleading. What I meant is that under the proposed scheme, users with existing v1 drivers must configure a v1 device explicitly. According to the plan, drivers will be updated so they can work with both v1 devices and new v2 devices. But if you configure a v2 device, old drivers will not work. -- MST